Why another output shape

Every synthesis tool eventually answers the same question: what should come out the other side?

Most chat interfaces punt on this. You ask, you get prose, you reshape it yourself. That reshaping is real work, and it happens every single time, even when the underlying request is nearly identical to last week's.

AirySyn starts from the output shape, not the conversation. A client update has a decision, a status, a next action, and a risk. A decision log has a decision, a reason, an owner, and an open question. Once the shape is fixed, synthesis becomes a much narrower problem: find the pieces that fill each slot, and say nothing else.

This is a smaller idea than it sounds. It does not require a better model. It requires being honest about what a useful result actually looks like, and refusing to ship anything less structured than that.